Satellite images geometric correction based on non-parametric algorithms and self-extracted GCPs

he geometric correction of high resolution satellite images can be carried out through generic non-parametric models that relates image to terrain coordinates. Traditional approaches to image geocoding relies on the measurement of a sufficient number of GCPs in both the ground and the image reference systems. Non-parametric models require a large number of GCPs well distributed on the whole scene, but the GCP identification and collection is a widely time-consuming operation and not always a simple task. Authors have developed two procedures for geometric correction based respectively on the Rational Function Model (RFM) and on a new neural network approach (MLP, Multi Layer Perceptron), and a procedure for automatic Ground Control Points (GCPs) extraction (AGE, Automatic GCPs Extraction) by means of a multi-resolution Least Squares Matching technique. This paper concerns about a new orthorectification procedure based on the sequential application of AGE, MLP and RFM algorithms for georeferencing high resolution satellite images. Tests have been carried out on Eros-A1 satellite images, using as reference maps available aerial orthoimages at a map scale of 1:10,000. A Case study is presented.